[dpdk-dev] [PATCH v5 14/52] common/cnxk: add npa bulk alloc/free support

Nithin Dabilpuram ndabilpuram at marvell.com
Tue Apr 6 16:41:06 CEST 2021


From: Ashwin Sekhar T K <asekhar at marvell.com>

Add APIs to alloc/free in bulk from NPA pool.

Signed-off-by: Ashwin Sekhar T K <asekhar at marvell.com>
---
 drivers/common/cnxk/roc_npa.h | 229 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 229 insertions(+)

diff --git a/drivers/common/cnxk/roc_npa.h b/drivers/common/cnxk/roc_npa.h
index 6983849..b829b23 100644
--- a/drivers/common/cnxk/roc_npa.h
+++ b/drivers/common/cnxk/roc_npa.h
@@ -8,6 +8,11 @@
 #define ROC_AURA_ID_MASK       (BIT_ULL(16) - 1)
 #define ROC_AURA_OP_LIMIT_MASK (BIT_ULL(36) - 1)
 
+/* 16 CASP instructions can be outstanding in CN9k, but we use only 15
+ * outstanding CASPs as we run out of registers.
+ */
+#define ROC_CN9K_NPA_BULK_ALLOC_MAX_PTRS 30
+
 /*
  * Generate 64bit handle to have optimized alloc and free aura operation.
  * 0 - ROC_AURA_ID_MASK for storing the aura_id.
@@ -141,6 +146,230 @@ roc_npa_aura_op_available(uint64_t aura_handle)
 		return reg & 0xFFFFFFFFF;
 }
 
+static inline void
+roc_npa_aura_op_bulk_free(uint64_t aura_handle, uint64_t const *buf,
+			  unsigned int num, const int fabs)
+{
+	unsigned int i;
+
+	for (i = 0; i < num; i++) {
+		const uint64_t inbuf = buf[i];
+
+		roc_npa_aura_op_free(aura_handle, fabs, inbuf);
+	}
+}
+
+static inline unsigned int
+roc_npa_aura_bulk_alloc(uint64_t aura_handle, uint64_t *buf, unsigned int num,
+			const int drop)
+{
+#if defined(__aarch64__)
+	uint64_t wdata = roc_npa_aura_handle_to_aura(aura_handle);
+	unsigned int i, count;
+	uint64_t addr;
+
+	if (drop)
+		wdata |= BIT_ULL(63); /* DROP */
+
+	addr = roc_npa_aura_handle_to_base(aura_handle) +
+	       NPA_LF_AURA_OP_ALLOCX(0);
+
+	switch (num) {
+	case 30:
+		asm volatile(
+			".cpu  generic+lse\n"
+			"mov v18.d[0], %[dst]\n"
+			"mov v18.d[1], %[loc]\n"
+			"mov v19.d[0], %[wdata]\n"
+			"mov v19.d[1], x30\n"
+			"mov v20.d[0], x24\n"
+			"mov v20.d[1], x25\n"
+			"mov v21.d[0], x26\n"
+			"mov v21.d[1], x27\n"
+			"mov v22.d[0], x28\n"
+			"mov v22.d[1], x29\n"
+			"mov x28, v19.d[0]\n"
+			"mov x29, v19.d[0]\n"
+			"mov x30, v18.d[1]\n"
+			"casp x0, x1, x28, x29, [x30]\n"
+			"casp x2, x3, x28, x29, [x30]\n"
+			"casp x4, x5, x28, x29, [x30]\n"
+			"casp x6, x7, x28, x29, [x30]\n"
+			"casp x8, x9, x28, x29, [x30]\n"
+			"casp x10, x11, x28, x29, [x30]\n"
+			"casp x12, x13, x28, x29, [x30]\n"
+			"casp x14, x15, x28, x29, [x30]\n"
+			"casp x16, x17, x28, x29, [x30]\n"
+			"casp x18, x19, x28, x29, [x30]\n"
+			"casp x20, x21, x28, x29, [x30]\n"
+			"casp x22, x23, x28, x29, [x30]\n"
+			"casp x24, x25, x28, x29, [x30]\n"
+			"casp x26, x27, x28, x29, [x30]\n"
+			"casp x28, x29, x28, x29, [x30]\n"
+			"mov x30, v18.d[0]\n"
+			"stp x0, x1, [x30]\n"
+			"stp x2, x3, [x30, #16]\n"
+			"stp x4, x5, [x30, #32]\n"
+			"stp x6, x7, [x30, #48]\n"
+			"stp x8, x9, [x30, #64]\n"
+			"stp x10, x11, [x30, #80]\n"
+			"stp x12, x13, [x30, #96]\n"
+			"stp x14, x15, [x30, #112]\n"
+			"stp x16, x17, [x30, #128]\n"
+			"stp x18, x19, [x30, #144]\n"
+			"stp x20, x21, [x30, #160]\n"
+			"stp x22, x23, [x30, #176]\n"
+			"stp x24, x25, [x30, #192]\n"
+			"stp x26, x27, [x30, #208]\n"
+			"stp x28, x29, [x30, #224]\n"
+			"mov %[dst], v18.d[0]\n"
+			"mov %[loc], v18.d[1]\n"
+			"mov %[wdata], v19.d[0]\n"
+			"mov x30, v19.d[1]\n"
+			"mov x24, v20.d[0]\n"
+			"mov x25, v20.d[1]\n"
+			"mov x26, v21.d[0]\n"
+			"mov x27, v21.d[1]\n"
+			"mov x28, v22.d[0]\n"
+			"mov x29, v22.d[1]\n"
+			:
+			: [wdata] "r"(wdata), [loc] "r"(addr), [dst] "r"(buf)
+			: "memory", "x0", "x1", "x2", "x3", "x4", "x5", "x6",
+			  "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14",
+			  "x15", "x16", "x17", "x18", "x19", "x20", "x21",
+			  "x22", "x23", "v18", "v19", "v20", "v21", "v22");
+		break;
+	case 16:
+		asm volatile(
+			".cpu  generic+lse\n"
+			"mov x16, %[wdata]\n"
+			"mov x17, %[wdata]\n"
+			"casp x0, x1, x16, x17, [%[loc]]\n"
+			"casp x2, x3, x16, x17, [%[loc]]\n"
+			"casp x4, x5, x16, x17, [%[loc]]\n"
+			"casp x6, x7, x16, x17, [%[loc]]\n"
+			"casp x8, x9, x16, x17, [%[loc]]\n"
+			"casp x10, x11, x16, x17, [%[loc]]\n"
+			"casp x12, x13, x16, x17, [%[loc]]\n"
+			"casp x14, x15, x16, x17, [%[loc]]\n"
+			"stp x0, x1, [%[dst]]\n"
+			"stp x2, x3, [%[dst], #16]\n"
+			"stp x4, x5, [%[dst], #32]\n"
+			"stp x6, x7, [%[dst], #48]\n"
+			"stp x8, x9, [%[dst], #64]\n"
+			"stp x10, x11, [%[dst], #80]\n"
+			"stp x12, x13, [%[dst], #96]\n"
+			"stp x14, x15, [%[dst], #112]\n"
+			:
+			: [wdata] "r" (wdata), [dst] "r" (buf), [loc] "r" (addr)
+			: "memory", "x0", "x1", "x2", "x3", "x4", "x5", "x6",
+			  "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14",
+			  "x15", "x16", "x17"
+		);
+		break;
+	case 8:
+		asm volatile(
+			".cpu  generic+lse\n"
+			"mov x16, %[wdata]\n"
+			"mov x17, %[wdata]\n"
+			"casp x0, x1, x16, x17, [%[loc]]\n"
+			"casp x2, x3, x16, x17, [%[loc]]\n"
+			"casp x4, x5, x16, x17, [%[loc]]\n"
+			"casp x6, x7, x16, x17, [%[loc]]\n"
+			"stp x0, x1, [%[dst]]\n"
+			"stp x2, x3, [%[dst], #16]\n"
+			"stp x4, x5, [%[dst], #32]\n"
+			"stp x6, x7, [%[dst], #48]\n"
+			:
+			: [wdata] "r" (wdata), [dst] "r" (buf), [loc] "r" (addr)
+			: "memory", "x0", "x1", "x2", "x3", "x4", "x5", "x6",
+			  "x7", "x16", "x17"
+		);
+		break;
+	case 4:
+		asm volatile(
+			".cpu  generic+lse\n"
+			"mov x16, %[wdata]\n"
+			"mov x17, %[wdata]\n"
+			"casp x0, x1, x16, x17, [%[loc]]\n"
+			"casp x2, x3, x16, x17, [%[loc]]\n"
+			"stp x0, x1, [%[dst]]\n"
+			"stp x2, x3, [%[dst], #16]\n"
+			:
+			: [wdata] "r" (wdata), [dst] "r" (buf), [loc] "r" (addr)
+			: "memory", "x0", "x1", "x2", "x3", "x16", "x17"
+		);
+		break;
+	case 2:
+		asm volatile(
+			".cpu  generic+lse\n"
+			"mov x16, %[wdata]\n"
+			"mov x17, %[wdata]\n"
+			"casp x0, x1, x16, x17, [%[loc]]\n"
+			"stp x0, x1, [%[dst]]\n"
+			:
+			: [wdata] "r" (wdata), [dst] "r" (buf), [loc] "r" (addr)
+			: "memory", "x0", "x1", "x16", "x17"
+		);
+		break;
+	case 1:
+		buf[0] = roc_npa_aura_op_alloc(aura_handle, drop);
+		return !!buf[0];
+	}
+
+	/* Pack the pointers */
+	for (i = 0, count = 0; i < num; i++)
+		if (buf[i])
+			buf[count++] = buf[i];
+
+	return count;
+#else
+	unsigned int i, count;
+
+	for (i = 0, count = 0; i < num; i++) {
+		buf[count] = roc_npa_aura_op_alloc(aura_handle, drop);
+		if (buf[count])
+			count++;
+	}
+
+	return count;
+#endif
+}
+
+static inline unsigned int
+roc_npa_aura_op_bulk_alloc(uint64_t aura_handle, uint64_t *buf,
+			   unsigned int num, const int drop, const int partial)
+{
+	unsigned int chunk, count, num_alloc;
+
+	count = 0;
+	while (num) {
+		chunk = (num >= ROC_CN9K_NPA_BULK_ALLOC_MAX_PTRS) ?
+				      ROC_CN9K_NPA_BULK_ALLOC_MAX_PTRS :
+				      plt_align32prevpow2(num);
+
+		num_alloc =
+			roc_npa_aura_bulk_alloc(aura_handle, buf, chunk, drop);
+
+		count += num_alloc;
+		buf += num_alloc;
+		num -= num_alloc;
+
+		if (unlikely(num_alloc != chunk))
+			break;
+	}
+
+	/* If the requested number of pointers was not allocated and if partial
+	 * alloc is not desired, then free allocated pointers.
+	 */
+	if (unlikely(num != 0 && !partial)) {
+		roc_npa_aura_op_bulk_free(aura_handle, buf - count, count, 1);
+		count = 0;
+	}
+
+	return count;
+}
+
 struct roc_npa {
 	struct plt_pci_device *pci_dev;
 
-- 
2.8.4



More information about the dev mailing list